Motion blur test

How to run the motion blur test

Put the display in the mode you actually use: its own resolution, the refresh rate you play or work at, and the overdrive or response time setting you normally leave it on. Close heavy tabs and apps so the browser has nothing else to do, and sit where you normally sit.

The patterns begin moving once the page has measured your refresh rate, which takes a second or two. Three lanes show the same object at the same speed. The full rate lane moves it on every frame. The half rate lane moves it on every second frame, twice as far each time, and the quarter rate lane on every fourth frame, four times as far. Because the speed is the same, the only difference is how often the picture changes, and that is what you compare.

1. Watch the dropped frames counter. While it stays at zero, the lanes are being drawn as intended and the comparison is fair.
2. Follow the object with your eyes, the way you would a character in a game. Blur that comes from the display holding each frame only appears when your eyes track something moving.
3. Then hold your eyes still on one spot and let the object pass. Blur from tracking mostly goes away and steps become visible; a trail behind the object does not go away, because that comes from the pixels.
4. Change the background and the speed. Trails show best where the change in brightness is largest, so try all three backgrounds: the black half of the object against the light one and the white half against the dark one are the biggest steps. The fastest speed makes every difference plain.
5. Press Full screen for the cleanest view, and Space to pause on a frame to look at edges. In full screen everything except a short hint disappears, and the hint fades after three seconds of no mouse movement.

Speed is in CSS pixels per second, the pixels a web page works in. The object moves a whole number of device pixels per frame, so it always lands on the panel's own grid, and the page tells you when that whole number means a slightly different speed from the one you picked.

How to read what you see

• Motion blur from the display holding the frame: the object looks smeared sideways by about the distance it moves between updates. Most LCD and OLED screens hold each frame lit until the next one arrives, and your eye keeps moving while the image stays put, so it smears across your vision. At 960 px/s on a 120 Hz display that is about 8 pixels, and it gets wider in the half and quarter rate lanes. This is normal, not a defect. The fine line groups on the white half go grey first, then the small text and the checkerboard; the widest lines survive longest.
• Ghosting: a faint copy or a fading trail behind the object, still visible after it has moved on, and present even when you hold your eyes still. It means pixels are slow to change. It is usually worst for changes to and from dark greys, and often worse on VA panels and on IPS panels with the response time set to a slow level.
• Overshoot, also called inverse ghosting: a bright or dark outline that follows the edges, a corona that isn't part of the object. The panel is driven past the colour it should reach and settles back. It comes from an overdrive or response time setting that is too aggressive; one step lower often clears it. It shows best on the grey background.
• Judder and hitching: the object pausing for a moment, in every lane at once. That is a dropped frame, and the counter shows it. The half and quarter rate lanes step by design, so compare their steadiness with the full rate lane, not with perfection.
• A much sharper object than the expected blur: if the full rate lane looks clearer than its step size allows, the display is probably switching its picture off between updates, with backlight strobing or black frame insertion. That reduces blur at the cost of some brightness, and it can flicker.

What the browser can and can't tell

A page can measure the rate it draws frames at and count the frames it missed. It does both here, from the times the browser reports for each animation frame, and the object is placed by counting frames rather than by reading the clock. If it followed the clock, every late frame would turn into a visible jump, and a lane meant to be smooth would stutter for a reason that has nothing to do with the display. Counting frames means a dropped frame shows as one extra refresh of holding still, which is what a dropped frame is.

The dropped frames counter is the number of refresh periods that passed without a new frame, judged against the median frame time. The first second after the test starts, or after a hidden tab comes back, is left out, because a browser that is still settling runs unevenly. So are the gaps when the tab was hidden or the page stopped drawing, since nothing was shown then. If the measured rate changes for good, for example because the window was dragged to another monitor, the patterns start over with steps for the new rate and the summary says so.

What a page cannot do is see the screen. It can't time how quickly a pixel changes, so it can't report a response time in milliseconds; can't tell ghosting from overshoot; and can't detect strobing or black frame insertion. Those are judged by eye or with a camera, and the summary lists them as not tested instead of guessing. The measured refresh rate is what the browser draws at, which normally matches the display's current mode; whether the display shows every frame it is sent takes a camera, like the frame skipping check in the refresh rate test.